Humanoid Robot Walking Trajectory Generation and Hybrid Control

Past three decades witnessed a growing interest in biped walking robots because of their advantageous use in the human environment. However, their control is challenging because of their many DOFs and nonlinearities in their dynamics. Various trajectory generation and walking control approaches ranging from open loop walking to systems with many sensors and feedback loops have been reported in the literature. The tuning of the parameters of reference gait is a common complication encountered. Another important problem of the walking control is the ground force interaction of the swinging leg at landing. This paper uses position references for the upper body. Optimization techniques are employed to obtain suitable leg joint torques for the supporting leg to track body reference trajectories. Locomotion is achieved by a swinging leg control scheme which is independent of the body reference position tracking. This method is based partly on gravity compensation and virtual potential fields applied to the swinging leg, and partly relies on position control. Soft landing of the leg can easily be handled with this scheme. 3D dynamics and ground interaction simulation techniques are employed for a 12-DOF biped robot to test the proposed method. The simulations indicate the applicability of the method in real implementations.